Scientists at Universita della Svizzera Italiana, University of Applied Science and Arts of Southern Switzerland, and University of Zurich have created a software that enables drones to automatically detect and then follow forest paths. Using new drones, people can be found and quickly rescued in mountain and forest areas.
Thousands of people die in mountain and forest areas every year. Emergency centers in Switzerland alone get more than 1,000 calls from lost hikers. Drones are capable of completing rescue task without needing assistance of teams. They are inexpensive and can be deployed on a large number, which will reduce risk of injury and response time.
Scroll down for video
A group of scientists at University of Zurich and Dalle Molle Institute for Artificial intelligence have designed a software to teach quadcopter to recognize and follow trails. It is a big step in the field of robotics and artificial intelligence, which this means that drones soon will be used in rescue teams to fast-forward the search.
Drone Flies Freely in Terrain
“While drones flying at high altitudes are already being used commercially, drones cannot yet fly freely in complex environments such as dense forests. In these environments, any little error may result in a crash, and robots need a powerful brain in order to make sense of the complex world around them,” said Davide Scaramuzza, a professor from University of Zurich.
The drone used by researchers observes environment with pair of cameras—same way mobile phones work. Instead of depending on sensors, drone uses a powerful algorithm to take images and recognize trails. If the trails are visible, software sends the drone to that direction. “Interpreting an image taken in a complex environment such as a forest is incredibly difficult for a computer,” said Alessandro Giusti from Dalle Molle.
Integration of Neural Application
Swiss team solved this problem using neural network, a computer capable of solving tasks from training examples—same as our brain. To gather information to train the algorithm, researchers hiked for hours and took more than 20 images of trails using cameras on their helmets. Their efforts paid off and neural network was able to find direction with 85% accuracy – human complete same task with 82% accuracy.
Scientific director at Dalle Molle Institute Juergen Schmidhuber says, “Our lab has been working on deep learning in neural networks since the early 1990s. Today I am happy to find our lab’s methods not only in numerous real-world applications such as speech recognition on smartphones but also in lightweight robots such as drones. Robotics will see an explosion of applications of deep neural networks in coming years.”
The team explains that a lot of work still have to be done to fully integrate drone in forest areas to search for missing people. Many challenges have to be overcome before this technology can be integrated into real-time environment. “One day robots will walk side by side with humans to make our lives better”, said Professor Scaramuzza. Small robots are versatile and this field is moving forward at a rapid pace. Now that drones are learning to understand and follow trails, we have to teach them to be able to recognize humans—research explained.